1 paper
Wentao Wang, Hengyu Zhong, Yunhan Jiang +2
As new evidence arrives, a sequence model must update what it remembers and how memory influences predictions. While Transformers incur computation and cache costs scaling with con…